Menu
Subscribe to Holyrood updates

Newsletter sign-up

Subscribe

Follow us

Scotland’s fortnightly political & current affairs magazine

Subscribe

Subscribe to Holyrood
Automation or AI? A simple guide to the terms, the differences and where they overlap

Image provided by EAS - AI generated

Partner content

Automation or AI? A simple guide to the terms, the differences and where they overlap

You’ve probably heard a confusing mix of words and phrases relating to AI lately and been left wondering what on earth the difference is, if any. Automation, RPA, AI, generative AI, agentic AI, agents, and intelligent automation are often used interchangeably in supplier pitches, in the press, and sometimes in official guidance, which can leave you unsure what you’re being asked to buy. If you’re deciding where public money goes, the last thing you need is confusion.

The Scottish Government’s AI guidance puts the starting point well: begin with the problem and use the simpler solution if it meets people’s needs. Knowing the difference helps you decide which one is simpler.

Automation

Pure automation follows rules. Robotic process automation, or RPA, is software that is programmed to carry out the steps a person would take on a computer, and when to take them: opening a case, checking a field against a list, copying details from one system to another, sending a standard letter. You set the rules, and it gives the same result every time.

It’s very good at high-volume, stable, rule-based work. It reads text from screens, forms, emails and digital documents, provided the information arrives in a predictable place and format. Free text, handwriting or a scan that varies from one to the next is where AI comes in. Automation can’t learn on its own. It does exactly what it’s told, making it highly reliable.

AI

AI interprets information.  It is trained on huge amounts of data so it can make sense of language and patterns that rules can’t cover. Obvious tools include Copilot and ChatGPT, but it covers a much wider range. It can read letters and emails to determine what each is about and how urgent it is. It can pull relevant details from documents that all look different, summarise a long history before a review, or draft a reply based on your policies and past answers. It can analyse thousands of records to find the duplicates and unusual cases that nobody has time to read through.

In each of these, a person checks the result before it goes anywhere because AI can still get it wrong at times. The Scottish Government’s guidance understandably says AI shouldn’t be used where decisions can’t be explained or reviewed.

The overlap

Intelligent automation is the name for the two working together. Rules do the repetitive work, and AI handles the steps that need reading or judgement. For example, an application arrives, AI reads the covering letter and picks out what’s relevant, the rules check eligibility and update the record, and a member of staff signs off the outcome.

In my experience, you can get great results with pure automation and layer AI in later on to handle the parts RPA couldn’t. Some processes won’t need both, and some won’t need either, because the best solution is changing how the work is done.

Agentic AI

You’ll also hear about AI agents. An agent is AI given a goal that chooses its own steps to reach it, rather than following a set route. Though it’s still early days, the idea has real promise. One thing to watch out for is software sold as an agent that's really just automation or a chatbot with a new name.

How to tell them apart

If someone could write down every step on a sheet of paper and the tool follows them, it’s automation. If it copes with things that vary, such as differently worded emails or documents with different layouts, without anyone writing a rule for each one, it’s AI.

If you’re being sold AI, ask these questions:

  1. What is the AI doing, and at which step? Look for specific answers here, not just “AI-powered throughout.”
  2. How does it show when it’s unsure, and who checks those cases?
  3. How has it been tested, and can you show it getting something wrong?

 

Talk to us

If you're weighing up where automation or AI could help, we're happy to talk it through. We specialise in helping public sector organisations save time and money with automation that delivers measurable results. Ask about our Automation Opportunity Assessment.  

Get in touch at info@easuk.co.uk

 

This article is sponsored by EAS

www.easuk.co.uk

 

 

Holyrood Newsletters

Holyrood provides comprehensive coverage of Scottish politics, offering award-winning reporting and analysis: Subscribe

Tags

Connect

Get award-winning journalism delivered straight to your inbox

Get award-winning journalism delivered straight to your inbox

Subscribe

Popular reads
Back to top